EleVint SimLab runs a resort the way a resort actually runs: someone asks for something, and the person on shift — the night auditor, the host, the cellar master — does the work on the platform: booking the room, opening the check, moving the wine, taking the payment. Shelby is the AI inside EleVint; when an employee is the one asking (a winemaker wanting Brix at 4 a.m.), it is Shelby who answers.
To prove it holds up, we let AI play both sides. Every staff member is a real employee on the books who clocked in for that shift, named on the conversation. An AI guest or colleague makes a realistic request, in their own words, mistakes and all, and it has to be satisfied by driving the real product — the same software our customers use, in real time. Nothing here is scripted or mocked up, and the tag on each conversation (Front Desk) is the shift being covered.
Open any conversation below to read exactly what happened, in order: the guest speaks, Shelby answers, Shelby goes to the system, and The system returned shows what actually came back before Shelby speaks again. Shelby went to the system lines are the real work carried out. Lines in red are where the system refused or fell short — we publish those too, because that list is our engineering queue.
A conversation with no Shelby went to the system lines is marked Never used the system. Shelby and the guest talked, but the product was never asked to do anything, so the conversation proves nothing either way. We count those separately rather than scoring them as passes.